Analysis of Variability in Car-Following Behavior Over Long-Term Driving Maneuvers

Abstract

The main goal of this dissertation has been to contribute to a better understanding of car-following behavior, and more specifically, on the variability in car-following behavior that is commonly observed in naturalistic driving situations.

This dissertation includes a thorough review of the literature in this area in which some important limitations of current car-following experimental studies and models, which make them inconsistent with naturalistic driving behavior under car-following situations, were investigated. A new data collection system using an instrumented test vehicle, with a synchronized user interface and data acquisition program coupled with two separate CAN networks, GPS, inertial distance measuring instrument, and digital video, has been developed to produce a sufficient quality and quantity of data on real driving behavior.

As a result of the data collection and analysis, we developed a better understanding of various behavioral characteristics in car-following behavior: (1) there was an oscillatory (or "drift") process in car-following behavior, which appears as a sequence of parabolic shapes in keeping desired following distance, (2) traffic hysteresis exists in car-following behavior, which is the phenomenon that drivers' acceleration and deceleration have different speed-density curves, (3) each individual driver has his or her own driving rule, rather than keeping a deterministic and strict driving law, but following distance for individual drivers can vary over time and space under different driving maneuvers and conditions, such as traffic, geometric, or environmental conditions, (4) drivers behave differently under different driving maneuvers, although they have exactly the same (current) instantaneous states, such as speeds of the lead and following vehicles and following distances, (5) reactions of following vehicle caused by the same driving maneuvers in car-following situations repeat themselves over time and space.

It was statistically evident from the analysis that different traffic and road characteristics (e.g., vehicle type, number of lanes, location of driving lane, and traffic condition), human characteristics (e.g., gender and distraction factors), and environmental characteristics (e.g., time of day and weather) have different effects on car-following behavior.

We hope that the findings of this dissertation will provide clues to guide the construction of more realistic car-following models to help improve the realism of microscopic traffic simulators, for which car-following logic is the core, and to develop more appropriate ACC algorithms and control strategies.